Repository navigation
ControlSurface v0.1.0-rc.1
Pre-releaseControlSurface v0.1.0-rc.1
First public release candidate of ControlSurface, an open-source production engineering platform for AI agents.
Highlights
- OTLP-native agent tracing with source spans and AgentRunGraph projections.
- Production Health and AI-specific SLO evaluation.
- Deterministic structural failure clustering.
- Incidents with inspectable Change Ledger evidence.
- Production-derived, human-reviewed regression cases.
- Paired candidate evaluation.
- Reproducible release gates with content-addressed evidence.
Watch the workflow
The 72-second launch film follows the deterministic refund-agent example from healthy traffic through a breaking tool-schema change, failure clustering, incident evidence, a representative trace, regression review, and a passing release gate. All application state shown comes from the running product. A 15-second preview and poster frame are attached. The original demo remains available for comparison.
Run it locally
Follow the Quickstart to start the supported Docker Compose deployment, create the owner, and send the first trace.
The refund-agent lifecycle demonstrates healthy traffic, a breaking tool-schema update, failure clustering, incident evidence, regression review, a fixed candidate, and a passing release gate.
Release status
This is an early source release candidate. Docker Compose is the supported V1 self-hosting path. The trust model is a trusted single owner on a trusted host; see Security and telemetry privacy and Operations.
Public Python-package and container-image distribution are intentionally deferred until the artifact-specific license bundle is complete. The tag includes source archives generated by GitHub; the revised launch film and its SHA-256 checksum are attached below.
Verification
The tagged commit passed the quality and hero-e2e checks, including clean-volume Compose boot, idempotent migrations, the real-service failure-to-release workflow, authenticated browser coverage, and login throttling. Detailed evidence is maintained in the release audit.